Wavelength division multiplex (WDM) networks offer high capacities and facilitate a wide variety of architectures. The finite system optical bandwidth and the effects of dispersion, however, restrict the overall number of channels carrying data and encourage small wavelength separation between the channels. Unfortunately, because the optical transmission fiber of an optical transmission system has a non-linear refractive index, four-wave mixing (FWM) penalties, also known as four-photon mixing penalties, can occur. As the wavelength separation between channels in a WDM network is designed to be smaller, the severity of FWM penalties associated with closely spaced channels correspondingly increases. This effect restricts the allowable wavelength separation of channels carrying data and, therefore, also limits the number of allowable channels within a WDM network.
Four-wave mixing occurs when two or more optical waves create a beat frequency whose oscillation modulates the refractive index of the optical fiber. This process generates sidebands that appear as noise, or intermodulation products (IMPs), in the neighboring channels. The magnitude of the noise for a given fiber is a function of the wavelength separation between interacting channels, the power in each channel, and the phase matching efficiency.
Several methods have been suggested to reduce or minimize FWM noise associated with closely located channels. For example, the phase matching, which is related to the magnitude of the FWM noise, can be decreased through the use of a high dispersion fiber (e.g., the commercially available fiber known as 1D). Although high dispersion within a fiber would decrease the phase matching and correspondingly decrease the FWM noise, high dispersion within a fiber imposes additional penalties such as pulse distortion over long distances.
Alternatively, because FWM noise is a function of the signal power in each channel, reducing the signal power or field amplitude for each channel has also been suggested. Because the power of the IMP generated by waves i, j, and k (P .sub.ijk) is proportional to the product of the individual component powers, P.sub.i P.sub.j P.sub.k, FWM noise can be greatly reduced by reducing field amplitudes. The signal power per channel, however, can only be reduced to a certain minimum level as dictated by the signal-to-noise ratio (SNR) limit for the system; if the signal power per channel is decreased below the minimal level, the signal power compared to the noise power may not be sufficient to maintain an acceptable low bit error rate (BER). For traditional fiber systems with non-return-to-zero (NRZ) modulation and optical amplification, the signal power for each channel typically cannot be made sufficiently low to avoid FWM penalties for low channel spacing. Thus, traditional fiber systems avoid FWM noise by increasing the wavelength separation between channels.
The above design modifications all suffer the same shortcoming: as one system parameter is adjusted to reduce FWM noise, other system parameters are affected which prevent the simultaneous reduction of wavelength separation between channels and the reduction of FWM noise. Thus, current design techniques limit the allowable minimum wavelength separation between channels.